Expected Annual Fraction of Entity Loss as a Metric for Data Storage Durability

Abstract

Data protection schemes are designed into storage systems to ensure data durability and accessibility in the presence of device failures. The effectiveness of these schemes has been evaluated based on the Mean Time to Data Loss (MTTDL) and the Expected Annual Fraction of Data Loss (EAFDL) metrics. The EAFDL metric assesses data loss at a low data processing unit level whereas durability refers to losses at a higher entity, say file or object, level. To evaluate the durability of storage systems we introduce the following reliability metric: the Expected Annual Fraction of Entity Loss (EAFEL), that is, the fraction of entities that is expected to be lost by the system annually. The general methodology that was applied to assess the MTTDL and EAFDL metrics is extended to obtain the EAFEL metric analytically for erasure-coding redundancy schemes and for the clustered, declustered, and symmetric data placement schemes. The theoretical model developed considers the effects of device failures, latent errors, and lazy rebuilds. For realistic values of sector error rates, the results obtained demonstrate that MTTDL and EAFEL degrade, but the EAFEL degradation is more pronounced when entities are large. It is also shown that the declustered data placement scheme offers superior reliability.